3D human pose estimation technology is used to transform 2D videos into 3D animated avatars that can be inserted in XR environments. In order to accomplish this task, appropriate deep learning algorithms are utilized to detect and track humans in the analyzed videos, extract their 2D poses, and then estimate their poses in the 3D space. The outcome is an animation of the humans’ movements that can be applied to selected 3D avatars.
3D human pose estimation can also assist other computer vision tasks such as action and emotion recognition. Thus, it enables a variety of real-life applications in domains such as Human–Computer Interaction, Video Surveillance, Biomechanics and Medication, Sports Performance Analysis and Education, Autonomous Driving, Psychology, and Try-on and Fashion.In the context of SUN, CERTH will use and expand state-of-the-art 3D pose estimation algorithms and adapt them so that they cover the needs of the specific use cases.